In modern urban areas jobs and residences are connected by commuting and by discretionary travel and consumers sort in land and labor markets according to idiosyncratic attachment, rents, wages and access to products. Our numerical simulations of a spatial general equilibrium model show that unpriced traffic congestion creates excess sprawl causing daily personal travel to be 8 min or 13% longer than optimal. We juxtapose congestion tolls, favored by economists, and urban boundaries, favored by planners, as two alternative policies for eliminating this excess. As explained in Anas and Rhee [Anas, A. and Rhee, H.-J. 2005, When are urban growth boundaries not second-best policies to congestion tolls?, in press in the Journal of Urban Economics], in our dispersed city, a boundary of any stringency is absolutely harmful and is not a second-best policy to congestion tolls. To curb the excess, the planner's boundary must be very stringent causing large distortions in land, labor and product markets while leaving the congestion unpriced. Such a boundary's deadweight loss is 15% of money income and 70 times higher than the benefits of firstbest congestion tolls that curb the sprawl by pricing travel efficiently. This result is robust in sensitivity analyses. Boundaries are efficient when compactness is valued, but tolls are still needed to reduce congestion.